4. FOUNDHERENTISM ARTICULATED The stimulation of his sensory receptors is all the evidence anybody has had to go on, ultimately, in arriving at his picture of the world. Quine ‘Epistemology Naturalized’! .there can be mutual reinforcement between an explanation and what it explains. Not only does a supposed truth gain credibility if we can think of something that would explain it, but also conversely: an expla- nation gains credibility if it accounts for something we suppose to be ‘rue, Quine and Ullian, The Web of Belief? l he goal is an explication of epistemic justification which conforms to the desiderata which emerged from the arguments of the pre- vious chapters: tovallow:the relevance of -experience.to, empirical justifica- tion (which will require an articulation of the interplay of causal and eval- uative aspects); and''to“allow" pétvasive-mutual. support. among. beliefs (which will require an account of the difference between legitimate mutual support and objectionable circularity). (Lhe explicandumis: Avis: more /less justified, at-time-t, in believing that prdepending.onw:.This choice of explicandum already indicates some sub- 117118 — EVIDENCE AND INQUIRY stantial presuppositions: that it is a:personal-Jocution, not an impersonal locution like ‘the belief that p is justified’ which is primitive; thatjustifica- tion:comes in degrees; that whether or to'what-degree’a person is justified in..believing something-may. vary. with~time. The rationale for these assumptions will become clearer as the articulation proceeds. My procedure might be called a ‘method of ‘successive approximation’, J begin with a formulation which seems intuitively very plausible but is (not surprisingly) also very vague, and attempt gradually to articulate gn ore precisely what is implicit in che intial, vague formula, The initial, © very approximate, first approximation is: Ais more/less justified)’at tj in Se believing that p, depending on-how good ‘his evidence is? I am inclined to regard this initial formulation as close to trivial (indeed, to think of ‘justi- fied’ as in effect epistemologists’ portmanteau word for what in ordinary parlance would most often be expressed in the less technical vocabulary of strong or flimsy reasons, a weak or overwhelming case, good or tenuous evidence, etc,). In the context of current debate in epistemology, however, it must be conceded that even this innocuous-seeming formula is not entirely innocent of presuppositions; it is, in particulary'an indication of a preference foran evidentialist over,an.extrinsic.approach,} For the present I shall offer no rationale for this preference over and above its intuitive plausibility. Later, however, this prima facie consideration will be but- tressed by arguments against extrinsic theories (chapter 7). Successive elaborations of the initial formula will depend on articu- lating the relations between the causal and the evaluative aspects of the concept of justification. Fundamental to this articulation will be the dis- tinction between the state and the content senses of ‘belief’, between someone's believing something and what they believe: a distinction marked in what follows as ‘S-belief’ versus ‘C-belief’* (henceforth, if I speak of h were ‘belief’, simpliciter, the ambiguity will be intended). How-justified:A:is.in ry y= that’ p’depends,"according-to-the-first-approximation,.on how x good-his. evidence is. Elaboration of »this first- approximation ‘will require * < three:stages: The first-stage, couched in terms-:of -causal:relations between A’sS-beliefs-and» other, including»perceptualy:states-of “A; will-be an attempt: to»characterize‘A's.-$-evidence.-with-respect-toup’»"The 'second, intermediate:stage will be’a manoeuvre by means"6f which to-arrive, on the»basis»of»-the«characterization-of “‘A’s S-evidence-with respect to p’ (which.consists of =certain»states~of “A)*at~a--characterization of, ‘A’sFoundherentism Articulated 119 G-evidencewith respect top’. (which. consists. of certain. sentences or propositions), The third; evaluative stage-will:complete the explication of ‘Aris more/lessjustified in believing:that p' by:characterizing ‘how good A’s C-evidence with respect.to-p is’. What is on offer will be at best a sketch of a theory—a sketch, fur- thermore, more than somewhat uneven in its level of detail. The reason, of course, is that this is, for the present at least, the best I can do. In hopes that I, or someone else, may find it possible eventually to improve the articula~ tion of the theory, I shall try to identify where the chief difficulties lie, and to discern which of those difficulties are peculiar to a foundherentist approach and which shared by more familiar theories of justification; though I shan’t make light of any of the difficulties, even when the prob- lems seem peculiar to my approach | shall take the attitude that those which arise only because my account is in some respects more detailed than its rivals should be regarded rather as challenges than as grounds for abandoning ship. I shall also try to be as clear as possible which parts of the formulation offered here are capable of standing independently, and may be usable, therefore, even if other parts fail. How justified a person is in believing something depends not just on what he believes, but on why he believes it, ‘why he believes it’ being a matter not simply of what else he believes, or of what else he believes and perceives, introspects or remembers, but of what it is, in his S-beliefs and experiences, on which his having the S-belief in question depends. (Consider two people both of whom believe that the accused is innocent, one because he saw her himself, a hundred miles away, at the time of the crime, the other because he thinks she has an honest face, The former is more justified than the latter.) Suppose, then, that A believes that p; how justified A is in believing that p depends in some fashion on what it is that causes him to have that S-belief. Asa first step towards spelling out ‘depends in some fashion on what itis that causes him to have that S-belief’ itisnecessary to distinguish the ini- tiating causes of “A’s:S-belief ‘that p— whatever. it was that. was involved.in his coming to believe that p originally—and the-causes-operative.at the.time in question, ie; at-the-time-at-which his-degree-of “justification is-at'issue,120 =~ EVIDENCE AND INQUIRY Roa These may be the same, but they may be different, and when they are dif- “s aw ya a oer Fe ferent, ivis:oi the:causes operativeat the time-in:question that justification depends. (Suppose that initially, at t;, A comes to believe that the accused is innocent for no better reason than that she has, he thinks, an honest face; but that later, at t, A learns that she has a watertight alibi, and it is this which, at that time, causes him to continue to believe she is innocent. He is more justified at ty than at ty) That is why the explicandum includes the condition, ‘at c’; this should, henceforth, be understood if it is not stated. As a second step, it is necessary to recognize that what causes someone to believe something, at a time, is often a matter of a balance of forces: some factors, that is, incline him towards believing that p, others incline him against it, with the former outweighing the latter. (Suppose that Prof. Smith believes that Tom Grabit stole the book, and that his S-belief is sustained by his remembering having seen Grabit leave the library surreptitiously with a guilty expression and a suspicious bulge under his sweater, and that this outweighs his desire not to believe ill of his students and his belief that it is possible, for all he knows, that Grabit has a light-fingered identical twin.) Socin:considering: what:causes“A»toshave'such-and“such ait S=belief, avty:itis necessary, to. distinguish sustaining'from inhibiting causes, Both, however, are'relevant to an‘assessment of ‘degree of justification. 4+ The third step is to distinguish those sustaining or inhibiting factors . }MSwhich are states of the person concerned from those which are not. (For example, A’s S-belief that there is a dog in the room might be sustained in part by his being in a certain perceptual state, and that state in turn be caused by there being a dog in the room.) Only:causes:of -A’s S-belief which‘are states of Awill'figure-in:the:characterization-of -his‘S-evidence. ‘The causal nexus, at t, of A’s S-belief that p’ will refer to those states of A which are operative at t, whether sustaining or inhibiting, in the vector of forces resulting in A’s believing that p. ‘Nexus’ is meant to sug- gest a mesh of S-beliefs interconnected with each other, with the subject’s perceptual experience, with his desires and fears, and so forth. The causal nexus of an S-belief is to include the states which directly sustain or inhibit that S-belief, the states which sustain or inhibit those states... and so on, The idea'is that our criteria of justification are neither simply atom- istic. nor unqualifiedly- holistic: they focus.on-those-elements.of.the whole constellation of .A’s states:at.t.which-bear-a. causal relation, sustaining or inhibiting,-to-the particular-S-belief -in-question.Foundherentism Articulated 121 Before even an initial explication of ‘A’s evidence with respect to p’ is possible, a-distinction. is-needed- between-evidential and non-evidential ged components within the causal nexus of an.S-belief. Belief ‘states, percep-ey" caval otateoy introspective Stites memoryzrncespuicclinside-evidential, S43 other states, such as the subject’s desires and fears, his being under the influence of alcohol or panic, etc., will not. That-such states.contribute to sustaining/inhibiting’A’s S+belief .that p may: have.a bearing on.the likeli- hood that p is true..(For instance, someone who is terribly afraid that p may turn out to be the case may greatly exaggerate the significance of his evi- dence that p is the case; ‘fearful thinking’, one might call this; someone who is under the influence of LSD is subject to extreme disorder of his senses; and so on.) Nevertheless, such components of the causal nexus of A’s S-belief that p are not counted as part of A’s evidence, because they are intuitively regarded as factors affecting a person’s reaction to, or judge- ment of, his evidence, not as themselves part of his evidence. That such non-evidential states belong to the causal nexus of an S-belief may well form a necessary part of an explanation of how it is that the subject believes something despite the flimsiness of his evidence; it will not, how- ever, form part of the calculation of the degree to which he is justified. ‘We now have the necessary apparatus for a preliminary conception of [gai ‘A’s evidence’, called, in an obvious extension of the state/content distinc- tion, ‘A's S-evidence with respect to p’5 ‘Ns S-reasons for believing that p’ |, j.4%~ will refer to those $-beliefs which sustain A’s S-belief that p; sory S-evidence for believing that p’ to the perceptual states which sustain A’s S-belief that p; ‘A’s past sensory S-evidence for believing that p’ to the perceptual traces which sustain A’s S-belief that p; ‘A’s sensory S-evidence for believing that p’ to A’s current and past sensory S-evidence for believing | that p; ‘A’s current introspective S-evidence for believing that p’ to the intro- spective states which sustain A’s S-belief that p; ‘A’s past introspective S-evi- dence for believing that p’ to the introspective traces which sustain A’s S- belief that p; ‘A’s introspective $-evidence for believing that p’ to A’s current and past introspective S-evidence for believing that p; ‘A’s experiential S-evi-} dence for believing that p’ to A’s sensory and introspective $-evidence fo: believing that p; and ‘A's S-evidence for believing that p’to A’s $-reasons and! experiential S-evidence for believing that p. ‘A’s S-evidence against believing that p’ will, be characterized like ‘A’s S-evidence for believing that p’, but with ‘inhibit’ for ‘sustain’; and ‘A’s S-evidence with respect to p’ will refer to122 EVIDENCE AND INQUIRY A’s S-evidence for believing that p and A’s $-evidence against believing that p.\A’s direct S-evidence with respect to p’ will refer to those evidential states which directly sustain/inhibit his S-belief that p, ‘A’s indirect S-evidence with respect to p’ will refer to those which directly sustain/inhibit his direct S-evidence with respect to p...and so on. A’s S=reasons-with:respect.to.p.are themselves’ S-beliefs.of »A’sj:-with srespect:to, which, Acmay-have further-S-evidence (which will be part of his $-evidence with respect to p). But A’s experiential-S-evidence-with respect to-p.consists of non-belief states of A,.not.the kind-of thing with respect to-which.A-has, or-needs,evidence..Experiential S-evidence evidentially sustains/inhibits S-beliefs, but not vice versa. A’s"experiential’S-evidence cont is, one-might say, his ultimate S-evidence>(This isthe important truth-that ae experientialist foundationalism:tries to accommodate—but in a forced and ‘unnatural.way.) ow The pre-analytic notion of ‘the evidence of the senses’ is not innocent of theory. Human beings, according to the commonsense picture, perceive things and events in the world around them; one interacts, by means of one’s senses, with the things in one’s surroundings; these interactions are what ‘sensory experience’ refers to. Our senses are, by and large, good at detecting what goes on around us; but in unfavourable circumstances one may be unable to see or hear clearly, and may misperceive, and in extremely unfavourable circumstances where one’s senses are grossly dis- ordered one may even ‘perceive’ what is not there at all. The previous sentence with its scare quotes indicates that the com- monsense conception takes for granted that ordinarily a subject's percep- tual states are the result of his sensory interactions with things around him, but that in extraordinary circumstances the subject could be in a state which is indistinguishable by him from the states resulting from his sensory interactions with the world, which are however wor the result of such inter- actions, but the product of some disorder in himself. The intention is to represent both the positive and the negative aspects of this picture. In-what follows; ‘perceptual:state’-will-be:given’ a'somewhat lax-interpretation,. to include ...states..phenomenologically-~indistinguishablé™*from perceptual states.in.the strictersense.,When it‘ comes to:the transition fromthe causal tothe” évaluative=stage “of ~the~explication;however,” the” commonsense assumption that perceptual states.are ordinarily the’result of “one’s sensory interactions’ with-things-and events ‘in the world will'be-built.in.Foundherentism Articulated 123 ‘Introspective S-evidence’ has been included as a kind of experiential S-evidence in the belief that it is also part of the commonsense picture underlying our pre-analytic conception of justification that a human being has some means of awareness of (some of) his own mental states and processes, as well as senses for scanning things and events in the world. But nothing will be said about introspection here beyond the observation that sensory-S-evidence.-and. introspective. S-evidence. are. treated as distinct’ 4. sl Sue with. the intention. of .avoiding any. confusion:of the: two,-any-elision-of ov” perception into introspective.awareness of.one’s.own-mental states. Such an elision would betray the commonsense presumption that what we per- ceive are the things around us—which I wish, on the contrary, to preserve. As the role played by perceptual states locates the relevance of cur- rent sensory experience to justification, the role played by perceptual [and introspective] traces locates the role of memory, in the sense represented by the locution ‘A remembers seeing/hearing/ etc...’ Here again the ter- minology will be used with deliberate laxity. ‘Perceptual [introspective] traces’ will be allowed possibly to include states which are indistinguish- able by the subject from those which are the present traces of past per- ceptual [introspective] states. ‘The distinction of perceptual state/perceptual trace, of current/past sensory S-evidence, is very crude—probably cruder than the pre-analytic ideas it represents. Perception isn’t instantaneous, but an ongoing process. But degree of justification can change in the course of the process, when, for instance, one gets a better look at a thing (‘It looked just as if someone was standing at the front door, until I got closer and saw it was only the shadow of the hydrangea bush’). To:mitigate the crudeness of the distinc~ tion of »current/past--sensory.-S-evidence..somewhat,..‘perceptual»-state’ should-be-construed:not.as-instantaneous, but as-having some.unspecified, gerrymanderable:duration. Past sensory S-evidence represents one of the ways in which memory fits into the picture. Memory also crops up in a second form: to say ‘A remembers that p’ is to say that he earlier came to believe that p and now still believes it, he has not forgotten it (and, of course, that p is true)-How justified -A-is-in such-a-‘persisting’ S-belief .will. depend, as-with all beliefs, on -how-good his.evidence—his evidence at the time-in. question—is-(This needn’t mean that it has to be said that I am not justified, for instance, in believing that my high-school English teacher's name was ‘Miss Wright’;124 — EVIDENCE AND INQUIRY this persisting belief is now sustained by past experiential S-evidence—of seeing and hearing the name used by myself and others, and so forth.) A person's S-beliefs are often maintained, in whole or part, by his hearing, seeing, or remembering hearing or seeing, what someone else says or writes. Suchetestimonial. evidence, as one might call it in an obvious extension of the usual sense, enters:the picture -by-way.of..the-rolesplayed by.A’s.sensory-S-evidence;-as-when-A’s-S-belief.that.p:is-sustained by-his remembering hearing-B say:that p, and-his-S-beliefsthavB is well-informed and:that.B-has.no.strong motive for deceit-or concealment on-this-matter. (It is assumed that if A doesn’t understand B's language, if he has the S- belief that p, his hearing B say "p’ will not form part of its causal nexus.) A's S-evidence with respect to p consists of a gerrymandered collection of [0% ge states of A, Butin-che evaluative: stage. ofthe-explication ‘evidence’ wil x + a qs” have:to:mean-'C-evidence’, for it is sentences or propositions, not states of * ey. 2 Person, which can support or undermine each other, probabilify or dis- ce confirm each other, be consistent or inconsistent with each other, cohere or fail to cohere as an explanatory story. So a bridge is needed from S- to C- Evidence. ‘A’s C-reasons for believing that p’ will refer to the C-beliefs A’s jeving that p; ‘NS experien- wo tial C-evidence for believing that p’ to sentences or propositions to the Cons, | effect that A is in a certain state or states—the state(s) which constitute(s oo yeu | A’s experiential S-evidence for believing that p; ‘A’s C-evidence for Ps g that p we believing that p’ will refer to A's C-reasons for believing that p and A’s expe- g that p g that p Ps oe sential C-evidence for believing that p; ‘A's C-evidence against believing ak hat p’ will be characterized like ‘A's C-evidence for believing that p’, but P g that p ith ‘against’ for ‘for’; and ‘A’s C-evidence with respect to p’ will refer to A’s evidence for believing that p and A’s C-evidence against believing that p. 's direct, indirect, indirect, etc., C-evidence with respect to p is distin- ished in parallel to the corresponding distinctions for A’s S-evidence. This»was:couched with deliberate:vagueness:in:terms.of.‘sentences or yor, propositions’. The main advantage of this deliberate vagueness is that, wv" because of the lack of clear criteria of identity for propositions, it tem- wy porarily puts off some hard questions, eg, about what sorts of character-Foundherentism Articulated 125 ization of perceptual (etc.) states might be appropriate here, Our ordinary ways of describing ‘the evidence of the senses’ offer some clues. What jus- tifies me in believing that there's a woodpecker in the oak tree-—‘my seeing it, the fact that I see it’, is a natural answer, an answer, however, often enough qualified or hedged, as: ‘but I only got a glimpse’, or ‘but it’s against the light’, or ‘but it’s too dark to see the markings clearly’, and so forth, and possibly revised to ‘well, it looked just as if there was a bird there’ It seems desirable to tie ‘A’s sensory evidence’ at least loosely to ‘how it looks (etc.) to As-and-ar-the same tine tO Fespect the commonsense distinction of more and less favourable circumstances—a good look is better evidence than a glimpse or glance, seeing a thing in full view and good light better evidence than a look at it partly hidden and at dusk ... and so on. For these ko (and other) reasons dvam»inclined:to:favour*characterizations“along:the We lines of ‘Ac is.in.the sort of perceptual state.a:normal:subject-would:bein, x” in-normal-circumstances,-when:looking-at-a rabbit three feet:away-and-in good light’;‘in the sort of :perceptual state a normal subject would be in;in, ve normal--circumstances,-when- getting a: brief “glimpse ofa fase-moving, ‘et rabbit at dusk’: and:so.on. Thisis:how, though ‘perceptual state’ has beer * ' con allowed to include states phenomenologically indistinguishable from those aN v 9 resulting from one’s. sensory. interactions. with-the. world; the presupposi- tion that normal perception is the result of such interactions is retaitied. There is another significant asymmetry built in, this time at the level of C-evidence, between A’s reasons and his experiential evidence. A’s:G- Feasons with respect to p will:consist of propositions which may-be true‘of may: be false. .His-experiential C-evidence, however, will. consist of sen teices-orspropositionsall-ofwhich-ace:trug, This is no reinstatement of & any sort of infallibilism with respect to perceptual or introspective beliefi;a\ it is just that the propositions concerned are to the effect that A is in ess and-such a perceptual (etc,) state, and they are all true because, ex bypotbesi, A is in that perceptual (etc:) state. This’ feature guarantees what may be’ called:the-‘experiential anchoring’ of .justified.empirical beliefs. Se How justified-someone-is.in believing:something, then, according toy) the second approximation,:depends:on: how: good his.C-evidencesis. Tie —_ remaining problem is to explicate ‘how good’. Before turning to that task, oN however, lest anyone, struck by the fact that the second approximation is couched entirely in terms of C-evidence, should suspect that the causal \” “pe aspect of the theory was redundant after all, it should be re-emphasized C/126 — EVIDENCE AND INQUIRY that the characterization of 's G-evidence'with respect to:p’:is'dependent on’the-characterization..of .A’s S-evidence with respect to p’,-which-the causal:part of the-theory-supplied. Which sentences or propositions con- stitute A’s C-evidence with respect to p depends on what states feature in the vector of forces maintaining A’s S-belief that p. ‘That justification comes in degrees is attested by numerous familiar locu- tions: ‘he has some justification for thinking that...’ ‘he would be more justified in thinking that...if...’; ‘his evidence is quite strong/at best flimsy/somewhat partial /one-sided’; ‘his grounds are reasonable/quite reasonable /overwhelming;; ‘his evidence gives colour to/gives some cre- dence to...’—Roget’s Thesaurus has an entire section entitled ‘Degrees of Evidence’, The explication being articulated here aspires to respect the gradational character of justification; not, however, to offer anything like a numerical scale of degrees of justification, or even anything as ambi- tious as criteria for a linear ordering, but only to say what factors raise, and what factors lower, the degree to which someone is justified in believing something. The model is not, as a foundationalist’s might be, how one determines the soundness or otherwise of a mathematical proof, it is, rather, how one determines the reasonableness or otherwise of entries in a crossword puzzle.$ This model is more hospitable to a gradational account, But the main motivation is that the crossword model permits pervasive mutual support, rather than, like the model of a mathematical proof, encouraging an essentially one-directional conception. The clues are the analogues of the subject’s experiential evidence; already filled-in entries, the analogue of his reasons. The clues don’t depend on the entries, but the entries are, in variable degree, interdependent, these are the analogues of the asym- metries already noted between experiential evidence and reasons. How reasonable one’s confidence is that a certain entry in a crossword puzzle is correct depends on: how much support is given to this entry by the clue and any intersecting entries that have already been filled in; how reasonable, independently of the entry in question, one’s confidence is that those other already filled-in entries are correct, and how many of theFoundherentism Articulated 127 intersecting entries have been filled in. Analogously, how good A’s C-evidence with respect to p is would depend on: (1) how favourable A’s direct C-evidence with respect to p is; (2) how secure A’s direct C-reasons with respect to p are, indepen- dently of the C-belief that p; (3) how comprebensive A's C-evidence with respect to p is. It should be noted that, although clause (2) mentions explicitly only A’s direct C-reasons with respect to p, its application takes one progres- sively outward, to the appraisal of A’s indirect, indirect)... etc, C-evidence with respect to p. For in considering how independently secure A’s direct C-reasons are, it will be necessary to consider how well his indirect; C-evidence supports them, and how independently secure his indirect, C-reasons are... and so on. C-evidence may be favourable or unfavourable with respect to a C-belief, with its being conclusive representing one extreme, and its pre- cluding the truth of the proposition in question representing the other. C-evidence may be favourable but not conclusive, supportive to a greater or lesser degree; or unfavourable but not fatal, undermining to a greater or lesser degree. One might say that at the upper limit evidence E makes it certain that p, at the lower limit E makes it certain that not-p; and that E is more supportive the more likely it makes it that p, the more undermining the more unlikely it makes it that p. But this, though true enough, is not very helpful, since ‘E makes it certain that p’, ‘E makes it likely that p’, and so forth are little more than verbal variants on the locutions in need of explication. One-might say; a'little more helpfully, that if-Eiis: conclusive it leaves:no‘room for alternatives to p, andif itis favourable but-not- conclu? sive itis the more supportive the less room it leaves for alternatives.to-p.I can’vresist calling this ‘the Petrocelli Principle’. With respect to the limit cases, I suggest the following rather straightforward characterization. E is conclusive with respect to p just in case its p-extrapolation (the result of adding p to it) is consistent, and its not-p-extrapolation inconsistent; E is fatal with respect to p just in case its not-p-extrapolation is consistent, and its p-extrapolation inconsistent. ‘The characterization of degrees of supportiveness less than conclu- siveness presents more difficulty. The Petrocelli Principle offers some128 ~~ EVIDENCE AND INQUIRY clues, but not enough, I think, to determine a unique solution. It directs us, at any rate, to look at the success of p relative to its competitors. So here is a tentative first move. A proposition C[p] is a competitor of p iff (i) given E, it precludes p, and (ii) the C[p]-extrapolation of E is better explanatorily integrated than E is. A strong characterization of support- iveness might run somewhat as follows: E is supportive to some degree with respect to p just in case the addition of p to it improves its explana- tory integration more than the addition of any of its competitors does. A weaker characterization would go, rather, along these lines. E is supportive to some degree with respect to p just in case the addition of p to it improves its explanatory integration; E is the more supportive with respect to p the more the addition of p to it improves its explanatory integration more than the more of its competitors do. The crossword analogy pulls one somewhat in the direction of the weaker characterization, which I am therefore inclined to favour—though not by a very large margin. I had formerly favoured the conjecture that E is supportive with respect to p just in case its p-extrapolation is better explanatorily in- tegrated than its not-p-extrapolation, and the more supportive the more its p-extrapolation is better explanatorily integrated than its not-p- extrapolation. But I no longer think this can be correct; the problem is that, if p is potentially explanatory of E or some component of E, it is not to be expected that not-p will be a rival potential explanans. (For this reason, the conjecture was also not well motivated by the crossword analogy.) This now-rejected characterization of supportiveness was prompted in part by its isomorphism with the characterization of conclusiveness. With either of the characterizations now on the table, I note, at least an analogy of structure can be sustained: conclusiveness is a matter of the superiority of P over its megation with respect to consistency with E; supportiveness is a matter of the superiority of p over its competitors with respect to the explanatory integration of E. The proposed characterization is not equivalent to more familiar accounts appealing to deductive implication and inductive support of p by E; and where it differs it has certain advantages. Although, if E is conclu- sive with respect to p, it deductively implies it, the converse is not true without exception. If E is itself inconsistent, E deductively implies p, but it does not qualify as conclusive with respect to p. If E is inconsistent, not only its not-p-extrapolation but also its p-extrapolation is inconsistent.Foundherentism Articulated 129 This upshot, that inconsistent evidence with respect to p is, as I shall say, indifferent, is surely more plausible than the foundationalist line that it is conclusive; and it is achieved without succumbing to the excessively stren- uous coherentist thesis that if there is any inconsistency in A’s belief-set, he is not justified in any of his beliefs. The intuition is much stronger that there is such a thing as favourable- but-not-conclusive evidence than it is that there is such as thing as ‘induc- tive implication’ or ‘inductive logic’—certainly if ‘inductive logic’ is taken to indicate relations susceptible of a purely syntactic characterization. My approach to ‘E is supportive (favourable but not conclusive) with respect to p’ has, from this point of view, at least the negative advantage of requiring no appeal to an ‘inductive logic’ which is prone to paradox at best, perhaps mythical at worst” Perhaps it also has a positive advantage. At least, by appealing to the notion of explanatory integration in the explication of supportiveness, foundherentism borrows some of the intuitive appeal of the notions of (on the foundationalist side) inference to the best explanation and (on the coherentist side) explanatory coherence. Like these more familiar notions, it should be construed as undemanding with respect to truth; ie, as requiring the truth neither of explicantia nor of explicanda. The notion of inference to the best explanation is both one-directional and optimizing in character; the notion of explanatory coherence has neither characteristic.’ So the explication tentatively proposed here is closer to the latter, coher- entist notion, since, first, explanatory integration is taken to be a property possessed in varying degrees by sets of propositions; and, second, because of my weak preference for the weaker characterization of supportiveness, the p-extrapolation of E does not have to be better explanatorily inte- grated than all C[p]-extrapolations for E to count as supportive with respect to p. How favourable E is with respect to p is not sufficient by itself to determine degree of justification. If A’s direct C-evidence with respect to p includes other beliefs of his, the degree to which he is justified in believing that p will also depend on the degree to which he is justified in believing those C-reasons. The possibility of mutual dependence is not precluded; it could be that A’s C-reasons with respect to p include some C- belief, say the C-belief that z, one of A’s C-reasons with respect to which is the C-belief that p. The point of the qualification ‘independently of the130 EVIDENCE AND INQUIRY C-belief that p’ in clause (2) is to avoid the danger of circularity this would otherwise present. The idea of independent security is easiest to grasp in the context of the crossword analogy, so I shall discuss it with reference to the small, one- spade crossword puzzle in figure 4.1. How reasonable one’s confidence is that 4 across is correct depends, inter alia, on how reasonable one’s confidence is that 2 down is correct. True, how reasonable one’s confidence is that 2 down is correct in turn depends, inter alia, on how reasonable one’s confidence is that 4 across is correct. But in judging how reasonable one’s confidence is that 4 across is correct one need not, for fear of getting into a vicious circle, ignore the support given it by 2 down; it is enough that one judge how rea- sonable one’s confidence is that 2 down is correct /eaving aside the support given it by 4 across, And this is just how my account of the independent security of A’s C-reasons with respect to p avoids vicious circularity. The crossword analogy also shows the way around another potential objection. The degree of independent security of A’s C-reasons with respect to p has been explained in terms of the degree to which A is jus- tified, independently of the C-belief that p, in believing his C-reasons with respect to p. So, since justified’ occurs on the right-hand side, won’t the explication be ineliminable? No—but the explanation is a bit tricky, and again easier to see in the case of the crossword puzzle. In figuring out how reasonable one’s confidence in some entry is, one will eventually reach a point where the issue is not how well some entry is supported by others, but how well it is supported by its clue. Analogously, in appraising how justified A is, independently of the C-belief that p, in believing his C-reasons with respect to that belief, one will eventually reach a point where the issue is not how well some belief is supported by other C-beliefs, but how well it is supported by experiential C-evidence. And the question of justification doesn’t arise with respect to experiential C-evidence.? But doesn’t this mean that the account is lapsing into a kind of foundationalism? No. What it means js that ‘justified’ eventually drops out of the explicans as one reaches the question, how well some belief(s) is (are) supported by experiential C-evidence; this does not require that any beliefs be justified exclusively by experiential C-evidence, nor, a fortiori, that 4 justified beliefs be justified by the support of such beliefs. (Recall the foundherentist interpretation of ‘the ultimate evidence for empirical beliefs is experience’.)Foundherentism Articulated 131 FIGURE 4.1 ACROSS DOWN 1. A cheerful start (3) 2. Angry Irish rebels (5) 4, She's a jewel (4) 3. Have a shot at an Olympic event (3) 6. No, it’s Polonius (3) 5. A measure of one’s back garden (4) 7. An article (2) 6. What's this all about? (2) 8. A visitor from outside 9. The printer hasn't got my number (2) fills this space (2) 9. What's the alternative? (2) 10. Dick Turpin did this to York; it wore ‘im out (5) Consider 4 across—RUBY Horo reasonable itis to tbink this is correct depends on: (1). the clue (2) how likely itis that IRATE is correct (3) how likely itis that PUT is correct (4) how likely it is that YARD is correct Horo reasonable itis to think that IRATE is correct depends on () the clue (i) how likely itis thet HIP is correct (which also depends on IRATE and PUT) (ii) how likely itis thet RAT is correct (which also depends on IRATE and RE) (iv) how likely itis that E'T is correct (which also depends on IRATE and RE) (v) how likely it is that ERODE is correct (which also-depends on IRATE, 00, and YARD) (vi) how likely itis that RUBY is correct How reasonable it is to think that PUT is correct depends on (a) the clue (b) how likely it is that HIP is correct (which also depends on IRATE and PUT) (c)_ how likely itis that RAT is correct (which also depends on IRATE and RE) (4) how likely it is that RUBY is correct How reasonable it is to think that VARD is correct depends on: (a) the clue (b) how likely it is that AN is correct (which also depends on YARD) (©) how likely it is that OR is correct (which also depends on YARD and OO) (@)_ how likely itis that ERODE is correct (which also depends on YARD, IRATE, and OO) (e)_ how likely it is that RUBY is correct132 EVIDENCE AND INQUIRY ‘There is an asymmetry to be noted between the role of A’s C-reasons for believing that p and the role of A’s C-reasons against believing that p. A is more [less] justified in believing that p the more [less] justified he is, independently of the C-belief that p, in believing his C-reasons for believing that p; but the less [more] justified in believing that p the more [less] justified he is, independently of the C-belief that p, in believing his C-reasons against believing that p. Degree of supportiveness and degree of independent security together are still insufficient to determine degree of justification; there remains the dimension of comprehensiveness. The comprehensiveness condition is the nearest analogue, in my account, of the more familiar total evidence requirement on inductions. Unlike this requirement, however, and like the comprehensiveness condition imposed by some coherentists, it is not a factor determining degree of supportiveness of evidence, but a separate criterion entering into the determination of degree of justification. Comprehensiveness promises to be tougher to spell out even than sup- portiveness and independent security; the crossword analogy isn’t much help here, and the characterization of ‘A’s evidence’ cannot be extrapolated in any easy way to ‘evidence’, simpliciter. Pethaps fortunately, the role of the comprehensiveness clause is most apparent negatively, when one judges someone unjustified or little justified in a belief because of their failure to take account of some relevant evidence. It is worth noting that ‘failure to take some relevant evidence into account’ includes failure to take a closer look, to check how the thing looks from the back, etc., etc; so the compre- hensivene: Even in advance of any further analysis, it is pretty clear that the condition must be construed to include experiential evidence. dimension of comprehensiveness is not likely to yield a linear ordering, And there is a further complication because the relevance of evidence is itself a matter of degree: an indeterminacy about how to weigh failure to take a lot of marginally relevant evidence into account relative to failure to take just a bit of more centrally relevant evidence into account.!° Relevance of evidence is being taken to be an objective matter. What evidence appears to A to be relevant depends on various background beliefs, which may be true or may be false. What evidence is relevant, however, coincides with what evidence appears to A to be relevant only if A’s background beliefs are true It can now be seen that inconsistency in one’s belief-set does carry a price, though a lesser price than the coherentist exacts. Inconsistency inFoundherentism Articulated 133 one’s C-evidence with respect to some belief has the consequence that one is not justified in that belief. To avoid this a subject whose beliefs are incon- sistent will have to keep the incompatible parts of his belief-set apart from each other; and this can be achieved only at the price of sometimes failing to take relevant evidence into account—which itself lowers the degree of justification of the beliefs it affects. ‘Acis more justified:in. believing that.p the more supportive-his:direct G-evidence: with: respect-to.p is, the-more. [less] independently, secure his direct.C-reasons for [against] believing that p.are,,and.the-more.compre- hensive his C-evidence with respect tops’. This is a bit more specific than the first and second attempts, but it still leaves the question, what the min- imal conditions are for A’s being justified to any degree in believing that p. One necessary condition is, simply, that there be such a thing as A’s C-evidence with respect to p; if his belief were the result of a blow to the head, for example, or of one of those pills philosophers are fond of imag- ining, he would not be justified at all. Furthermore, since it is the justifica_ tion_of empirical beliefs_which_is_at issue, it_is_necessary that_A’s C-evidence include some experiential C-evidence. (This is my analogue of BonJour’s Observation Requirement—note, however, that while his requirement was quite out of place in BonJour's coherentist theory, mine is precisely iv place in my experientialist foundherentism.) Another neces- sary condition has already been suggested, in the context of the discussion of supportiveness: A’s C-evidence must be favourable with respect to p. Presumably also some minimal standard omprehensiveness is neces- ary; it is tempting to suggest that ‘As C-evidence must at least include all the relevant evidence A possesses—but this is unfortunately too demanding. Given A’s other beliefs, some of the relevant evidence he pos- sesses may not appear relevant to A; worse, this suggestion would have the undesired consequence that, after all, any inconsistency in A’s belief-set precludes his being justified in any belief. With respect, finally, to the ques- tion of the minimal standards of independent security, the obvious sug- gestion is that A must be justified to some degree in believing his direct C-reasons for believing that p; but the asymmetry between A’s C-reasons for believing that p and his C-reasons against believing that p means that no such obvious suggestion offers itself on the negative side What about the upper end of the scale? Our ordinary talk of someone’s being ‘completely justified’ in believing something is highly134 EVIDENCE AND INQUIRY context-dependent; it means something like: ‘in the circumstances— including such matters as how important it is to be right about whether p, whether it is A’s particular business to know whether p, etc., etc—A’s evi- dence is good enough (supportive enough, comprehensive enough, secure enough) that he doesn’t count as having been epistemically negligent, or as epistemically blameworthy, in believing that p’. This may be represented by ‘A is completely justified in believing that p’, which will refer to a context- dependent area somewhere vaguely in the upper range of the scale of jus- tification. Its vagueness and context-dependence is what makes this ordi- nary conception useful for practical purposes (and for the statement of Grettier-type paradoxes). But philosophical talk of ‘complete justification’ is best construed in a more demanding, context-neutralized fashion. This may be represented by ‘A is COMPLETELY justified in believing that p’, which would require A’s C-evidence to be conclusive and maximally com- prehensive, and his C-reasons to be maximally independently secure. Now for some extrapolations and applications. Iv The explication taken as primitive here, ‘A is more/less justified in believing that p’, presupposes that A believes that p; but what of ‘A would be more/less justified in believing that p’, which presupposes that A doesn’t believe that p? The degree to which A would be justified, presumably, depends on how good the evidence is that A has with respect to p. So an explication of this locution would refer to those beliefs and experiences of A’s which are relevant to p, and to how justified A is in believing that p on the assumption that those belief and experiences are his evidence with respect to p, in the sense explained above. What about the possibility of speaking of the degree to which, not an individual, but a group of people, is justified in believing that p? (I am thinking, for instance, of the kind of case where a group of scientists pro- duces a report of work which is joint in the sense that different members of the group have done different but related parts of it, in more or less complete awareness of the others’ work.)!! It might be feasible to make some kind of sense of this by starting with the degree to which a hypo- thetical subject whose evidence includes all the evidence of each memberFoundherentism Articulated 135 of the group would be justified in believing that p, and then discounting this by some index of the average degree to which members of the group are justified in believing that the other members are reliable. If the result of pooling the evidence of the various members of the group were incon- sistent, this would have the consequence that the group, gua group, is not justified to any degree, even if some or even all its members were, in believing that p. This consequence seems correct. As for the impersonal locution, ‘the belief that p is more/less justified’, though I would not go so far as to say that no sense can be made of it, I have to say that I have no explication to offer. The problem may be in part that the locution works differently in different contexts; in some, perhaps, it may mean ‘someone is or would be more/less justified in believing that p’, but this doesn’t seem plausible as an all-purpose explication. The obstacle, of course, is that since empirical justification depends ultimately on experience, and since it is persons who have experience, an impersonal locution is prima facie out of place. It is possible to adapt the account suggested so as to accommodate the idea that belief, as well as justification, comes in degrees. (A gradational conception of belief is not obligatory, since there is the alternative of allowing degrees of approximation to belief construed categorically; but it is one way, and quite a useful one, of acknowledging that a person's acceptance of a proposition as true may be more or less complete.) The basic principle of adaptation is simple: that degree of justification is inversely related to degree of belief—ie,, assuming A’s evidence held con- stant, the less strongly A believes that p, the more justified he is in this (weak) belief. This principle is the nearest analogue, in my account, of Hume's injunction that one proportion one’s beliefs to the strength of one’s evidence.!2 It may look, however, as if there is an awkward compli- cation concerning A’s reasons with respect to a belief: how is the possibility that A believes his reasons less than fully to be accommodated? It turns out, however, that one can let the degree to which A is justified in believing that p depend, inter alia, on the degree to which he would be justified in fully believing his reasons with respect to p (even if, in fact, he only partially believes them). If A less-than fully believes his reasons for believing that p, either this will be reflected in a lowering of his degree of belief that p, or not. If so, the inverse relation between degree of belief and degree of justification already suggested will raise the degree to which he is justified136 EVIDENCE AND INQUIRY in his (properly weak) belief that p. If not, it will lower the degree to which he is justified in his (improperly strong) belief that p. And this— plus the usual asymmetrical mutatis mutandis clause for his reasons against believing that p—seems to be all that is needed. Let me close this section by seeing how my account fares with regard to the lottery paradox—a challenge for any theory of justification, and one which it is plausible to think a gradational account might have some advan- tages in meeting, Suppose A believes that ticket no. 1 won't win, and that his evidence is: that there are 1 million tickets, that exactly one ticket will win, that the chance that ticket no. 1 will win is 1/1 million, that the chance that ticket no. 2 will win is 1/1 million... etc, [E]. Suppose also that his evidence is com- prehensive, and his reasons perfectly secure. E is highly supportive but not conclusive with respect to the belief that ticket no. 1 won't win. Thatit is not conclusive follows from the explication given. It would take more articula- tion of the explication of supportiveness to establish conclusively that it is highly supportive, but this is precisely in line with the Petrocelli Principle, which is intended as a guide to further articulation, So A is justified to a high degree, but not COMPLETELY, in believing that ticket no. 1 won't win. The same argument applies if we suppose that A believes that ticket no. 2 won't win, and that his evidence is, again, E,...and so on for the belief that ticket no. 3 won't win...and the belief that ticket no. 1 million won't win. Now suppose that A believes that neither ticket no. 1 nor ticket no. 2 will win. If his evidence is, again, E, then it is, again, not conclusive; it is still highly supportive, but less supportive than it is of ‘ticket no. 1 won't win’ or of ‘ticket no, 2 won’t win’. So he would be justified, not COM- PLETELY, but to a high degree, though lower than the degree to which he would be justified in believing that ticket no. 1 won't win or in believing that ticket no. 2 won't win, in believing that neither ticket will win. If, now, one supposes that A believes that none of tickets nos 1-3 will win, again on evidence E, the same line of argument shows that he would be justified in believing this, but to a degree lower than that to which he would be jus- tified in believing that neither ticket no. 1 nor ticket no. 2 will win, As more conjuncts are added, A’s degree of justification would fall; he would be less justified in believing that none of tickets nos 1-100 will win, less justified again in believing that none of tickets nos 1-1,000 will win...and so on. In due course, E will cease to be supportve and becomeFoundherentism Articulated 137 undermining; A would not be justified in believing that none of tickets nos 1-500,001 will win. Now suppose that A believes that no ticket will win. If his evidence is E, it is not just not conclusive, not just not supportive, but fatal; for it includes: some ticket will win. So A would not be justified in believing that no ticket will win. But this, it may be said, misses the point. By assuming, throughout, that A’s evidence is just E, I have evaded the problem, which arises from the intuition that A would be justified in believing that ticket no. 1 won't win, justified in believing that ticket no. 2 won't win ... and justified in believing that ticket no. 1 million won't win, but vor justified in believing their con- junction, i.e. that no ticket will win. Very well. Go back to the case where A believes that neither ticket no. 1 nor ticket no. 2 will win, but now assume that his evidence is E, plus the C-belief that ticket no. 1 won’t win, plus the C-belief that ticket no. 2 won't win. Call this E’. E’ is conclusive with respect to the C-belief that neither ticket no. 1 nor ticket no. 2 will win. But though ex Aypothesi A is COM- PLETELY justified in believing E, he is not COMPLETELY justified in believing that ticket no. 1 won't win, nor in believing that ticket no. 2 won't win, So now, though his reasons are conclusive, they are not fully indepen- dently secure. It seems reasonable to asume (though this goes beyond what the explication of independent security is precise enough decisively to entail) that though he would be justified to a high degree in believing that neither ticket no. 1 nor ticket no. 2 will win, he would be justified to a lesser degree in believing the conjunction than he would be in believing either conjunct. And if so, then, as before, as more conjuncts are added, his degree of justification would fall. Finally, suppose that A believes that no ticket will win, and now that his evidence is E, plus the C-belief that ticket no. 1. won't win, plus the C- belief that ticket no. 2 won't win... plus the C-belief that ticket no. 1 mil- lion won't win [E*]. E* deductively implies that no ticket will win, But this is not sufficient to make it conclusive. In fact, E* is inconsistent (for it includes E, which includes the C-belief that exactly one ticket will win), and hence, by my account, indifferent. So, once again, A is not justified to any degree in believing that no ticket will win T hope this draws the sting of the paradox. It is more oblique than more familiar proposed solutions, but this is not necessarily a disadvantage.188 EVIDENCE AND INQUIRY Instead of simply requiring us to drop the principle that, if A is justified in believing that p and justified in believing that q, then he is justified in believing that p and q, it supplies the beginnings of an explanation of what is wrong with it. Briefly and roughly: it shows how it can be that A is justi- fied to some degree but not COMPLETELY in believing that p, and justi- fied to some degree but not COMPLETELY in believing that q, and his evidence with respect to p and q can be less good than his evidence with respect to p or his evidence with respect to q, or, indeed, no good at all. It may be helpful to back up this rather abstract observation with a homely and non-paradoxical example which points in the same direction. Suppose that A is justified to a high degree, but not COMPLETELY, in believing that colleague number 1 will be at tomorrow’s departmental meeting (his direct C-evidence being, say, that colleague number 1 is reli- able and conscientious and has never missed a meeting yet); and that A is justified to a high degree, but not COMPLETELY, in believing that col- league number 2 will be at the meeting (his direct C-evidence being, say, that colleague number 2 has repeatedly said that he is very keen to see that a certain item on the agenda is passed, and would not except in the direst emergency fail to turn up to vote for it). It seems intuitively clear that A is fess justified in believing that colleague number 1 will turn up and col- league number 2 will turn up. Why? Well, still speaking intuitively, it is because there is more to go wrong, more of a gap between his C-evidence and the truth of the conjunction. His C-evidence with respect to the belief that colleague number | will be there is not conclusive—A doesn’t know, eg, that there won't be a domestic emergency or car breakdown pre- venting him from showing up; and neither is his C-evidence with respect to the belief that colleague number 2 will turn up—A doesn’t know, eg, that colleague number 2 hasn't realized, overnight, that the proposed changes would involve a significant increase in teaching loads, and decided that the least embarrassing course of action is to invent an excuse for not showing up after all. And if any of these gaps in A’s C-evidence should turn out differently, one of the two colleagues would be absent, Complex as this has been, it is very far from complete. I have helped myself to a whole slew of concepts, some of which have been left completely unexplicated, and of none of which a fully satisfactory account has been given. This cannot be excused by appeal to the fact that the pre-analyticFoundherentism Articulated 139 conception of justification is itself vague; for one purpose of explication is to improve precision. Nor can it be excused by appeal to the fact that any explication must come to an end somewhere; for the concepts on which I am relying are hardly so transparent as to be prime candidates for this status. In partial mitigation, however, it may be observed that close rela~ tives of some of the concepts needed (explanatory integration, compre- hensiveness) are already current in the literature, and that the foundher- entist is free to borrow the best efforts of rival theorists to spell them out. ‘The successive approximations by means of which the explication has been presented mean that it would be possible to follow me through some steps without following me all the way. This is reassuring, since the suc- cessive steps get dicier as, and to the extent that, they get more specific. It would also be possible to borrow the double-aspect approach without sub- scribing to the foundherentist structure offered; to adapt the crossword model of the structure of evidential support without adopting the double- aspect approach; or to follow me with regard to the double-aspect approach and the crossword model without accepting my explication of supportiveness in terms of explanatory integration; and so on. But my hope is that this explication, imperfect as it is, represents at least approximately the kind of theory which, as I argued in chapters 1-3, is required to overcome the difficulties faced, on the one hand, by foundation- alism and, on the other, by coherentism. Still, there is quite a way yet to go. Since a prime motivation for seeking a ‘third alternative’ theory was, from the beginning, that a coherentist account cannot allow the relevance of experience to justification at all, while a foundationalist account can allow it only in a forced and unnatural fashion, it will certainly be neces- sary to say more than I have so far about the role of experience in the foundherentist account, It will be not only convenient, but also, I hope, illuminating, to combine this task with a critique of Popperian ‘episte- mology without a knowing subject’, which throws the problem of the role of experience into particularly sharp relief,